Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007

For Evaluation Only.

A review of islanding detection techniques for renewable distributed

generation systems
Aziah Khamis a,b,n, Hussain Shareef a, Erdal Bizkevelci c, Tamer Khatib a
a

Department of Electrical, Electronic and Systems Engineering, Universiti Kebangsaan Malaysia, Malaysia
Faculty of Electrical Engineering, Universiti Teknikal Malaysia Melaka, Malaysia
c
Power Systems Department, Tubitak Uzay, Ankara, Turkey
b

art ic l e i nf o

Keywords:
Islanding detection
Renewable distributed generation
Distributed generation (DG)

a b s t r a c t
Islanding detection of distributed generations (DGs) is one of the most important aspects of interconnecting DGs to the distribution system. Islanding detection techniques can generally be classied as
remote methods, which are associated with islanding detection on the utility sides, and local methods,
which are associated with islanding detection on the DG side. This paper presents a survey of various
islanding detection techniques and their advantages and disadvantages. The paper focused on islanding
detection using a conventional and intelligent technique. A summary table that compares and contrasts
the existing methods is also presented.

Contents
1.
2.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Islanding detection methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.
Central (remote) techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.1.
System state monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.2.
Switch state monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.3.
Intertripping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.
Local islanding detection techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1.
Passive techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.2.
Active techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.3.
Hybrid technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.
Signal processing methods used for identication
of island mode operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.1.
Wavelet-transform (WT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.2.
S-transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4.
Intelligent classiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4.1.
Articial neural network (ANN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4.2.
Probabilistic neural network (PNN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4.3.
Decision trees (DT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4.4.
Fuzzy logic (FL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Acknowledgement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
1. Introduction
Conventional power distribution systems are passive networks,
where electrical energy at the distribution level is always supplied
to the customer from upstream power resources that are connected to the bulk transmission system. The distributed generation
(DG) concept is introduced at the distribution level to exploit the
benets of small local renewable generation. The resources are
generally below a couple of MWs and can be wind farms, micro
hydro turbines, photovoltaics (PV), and other generators that are
supplied with biomass or geothermal energies. The difference
between traditional and embedded distribution network systems
is illustrated in Fig. 1. In an embedded distribution network
system, additional DG resources are supplied near the local load
compared with the traditional network system.
The application of multiple DGs in the distribution system is
becoming a common practice with the integration of DG resource.
This practice is caused by the advantages of DG such as environmental benets, increased efciency, avoidance of transmission
and distribution (T&D) capacity upgrades, and reduced T&D line
losses [15]. However, numerous problems should be tackled
before the DG units are applied to the networks. These problems
include frequency stabilization, voltage stabilization, intermittency
of the renewable resources, and power quality issues. The formation of the microgrid (MG), which is cause by the disconnection
from the main grid without stopping the energy generation from
the DG sources, can also be considered as a drawback of DG [6].
The disconnection of the main source is called islanding, which
can be either intentional or unintentional. The purpose of intentional islanding is to construct a power island during system
disturbances, which are commonly introduced because of the
faults. However, the active part of the distribution system should
sense the disconnection from the main grid and shut down the
distributed generators in countries where island mode operation
(MG) is not allowed. Undetected island MG is generally called
unintentional islanding. Fig. 2 shows the formation of a power
island because of an upstream fault in the grid system [7,8].
The unintentional islanding of DGs may lead to several problems in terms of power quality, safety, voltage and frequency
stability, and interference [35,911]. The IEEE 15472003 standard species a maximum delay of 2 s for the detection of the
unintentional islanding condition; the IEEE 9291988 standard
requires the disconnection of the DG if islanded [3,4]. Therefore,
uncovering effective solutions to resolve this problem is necessary.
Research work on unintentional islanding detection is rapidly
growing to ensure that the system is operated under the standard
requirements. The literature shows several technical publications
related to islanding detection from the past twelve years. Most of
the ideas that aimed to resolve the problem were proposed after
2007, which indicates the importance of the research subject.

2. Islanding detection methods

Various techniques have been developed to detect islanding.
These techniques can be broadly classied into central (remote)
and local methods as illustrated in Fig. 3. In the following
subsections, the details of these methods are explained and
evaluated.
2.1. Central (remote) techniques
2.1.1. System state monitoring
System state monitoring is a method for determining system
states from a model of the power system network with a reduced
number of state measurements. This method is generally regarded

consumers

consumers

consumers
consumers

consumers

consumers

consumers

DG
consumers

consumers
consumers
consumers

consumers

DG

DG
DG
consumers

consumers

Fig. 1. (a) Traditional distribution system (Traditional Grid System), (b) Generation
embedded (Microgrid Networks).
Tripped Utility Circuit
Breaker

Network
Load

Local
Load
DG

Utility
Power
Island

Industrial Site

Fig. 2. Illustration of power islanding condition.

as a function of the Distribution Management System (DMS),

which is a complementary of supervisory control and data
acquisition (SCADA) systems. The method is also used to detect
unintentional islanding by monitoring the parameters of the entire
distribution system such as voltage and frequency [12]. If the
parameter can still be detected from the disconnected area, the
occurrence of islanding is detected. This method is highly effective
in detecting unintentional islanding if the system is properly
instrumented and controlled. However, the cost of implementation is expensive because each inverter requires separate instrumentation and communication equipment. The survey shows that
this technique was tested by the PV system. Therefore, other DG
types, such as wind turbines and fuel cells, can be explored. The
limitation of the high cost of implementation, particularly for
small systems, can be addressed using other techniques.
In [6,10], the voltage sensitive devices embedded in the PV-based
DG inverter are connected to SCADA system. The loss of mains is
detected and notied to the central control system to inform the
island mode operation. Real time monitoring of voltage for each
generator in the distribution grid can be a cumbersome process with
an increased number of DGs connected to the grid. SCADA is also
used to monitor auxiliary contact on all circuit breakers between the
main source of generation and the DG units [13].

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
ISLANDING

CENTRAL CONTROL
(REMOTES)

LOCAL (RESIDENCES)

CONVENTIONAL

PASSIVE

STATE MONITORING

ACTIVE

SWITCH MONITORING

HYBRID

INTERTRIPPING

ARTIFICIAL INTELLIGENT

Fig. 3. The classication of islanding detection techniques.

Table 1
Summary of remote islanding detection techniques.
Methods
SCADA system

Advantages

 Communicate
with all DG

Disadvantages

Improvements

 Cost implementation high

 Difcult to x

Transfer trip
scheme

 Simple concept
 Avoid NDZ

 Expensive
 Complicated

 Direct transfer trip can avoid islanding

Wired cable or
non-wired

 Easy to be

 Cost concern for DG below 2 MW (cost impact can goes

 Needs any possible media to transfer the signals

 Concern in remote intertrip signals, where the communication

implement

up to 50% of total connection cost)

media is outside the control of the user

2.1.2. Switch state monitoring

The SCADA system can be used to monitor the status of the
circuit breakers and reclosers that could island a distribution
system [3,5]. However, this method requires an improved interaction between the utility and DG units, which leads to extra costs
for both utility and DG sides. Transfer trip detection schemes
require all the circuit breakers that island the DG to be monitored
and linked directly to the DG control, or through a central
substation SCADA system. When a disconnection is detected at
the substation, the transfer trip system determines which areas
are islanded and sends an appropriate signal to the DGs to remain
in or discontinue operation.
The transfer trip scheme is incorporated with SCADA to
monitor the status of circuit breakers and reclosers [313]. The
scheme allows for the additional control of DGs through the utility
and increases the coordination between the DGs and utility.
However, the method has a complexity cost because of the growth
of the system complexity, where the transfer trip becomes
outdated and requires relocation or updates.

2.1.3. Intertripping
Another method that can be used for islanding detection is
intertripping, which is theoretically different from central control
techniques. The method detects the opening of a contact at the
points of disconnection and transmits the signal to all generation
sites that support the respective island zones. Intertripping also
generally relies on the communication between the sensors and
generating units. These techniques have higher reliability and
provide accurate solutions but are uneconomical [14,15].
The above three techniques are used because of their reliability.
The review shows that central control techniques are preferable
because these techniques can avoid non-detection zones (NDZ),
where the power absorbed by the load almost perfectly matches
the power generated by the DG [10,16,17].

Central control methods are also not inuenced by the number

of inverters, size of system, type of generator, and penetration
level. However, a tight connection between utility and DG units is
needed in applying communication techniques. Table 1 shows a
summary of the remote islanding detection techniques.
2.2. Local islanding detection techniques
Local techniques are broadly used to detect islanding based on
the measurement of the system parameters at the DG site such as
voltage, frequency, current, and harmonic distortion. These techniques can be further divided into passive, active, and hybrid
techniques. A literature overview on local islanding techniques
was conducted. The statistic notably shows that active and passive
techniques have rapidly increase throughout the years.

2.2.1. Passive techniques

Passive techniques monitor the system parameters such as
voltage, frequency, harmonic distortion, and current on the DG site
at the point of common coupling (PCC) with the utility grid. These
parameters vary greatly when the distribution system is islanded.
The parameters typically used to detect islanding conditions are
frequency and voltage [5,7,1822]. Various traditional passive
islanding detection techniques exist as follows:

2.2.1.1. Under/over voltage and under/over frequency. The under/

over voltage (UVP/OVP) and under/over frequency (UFP/OFP) is the
oldest technique adapted to protection the distribution system. The
protection relays for this technique are placed on a distribution feeder
to determine the various types of abnormal conditions. UVP/OVP and
UFP/OFP are used to monitor of the grid voltage/frequency exist the
limits imposed by the relevant standards [10]. These technique is
usually used at all grid connected PV inverters. These protection
methods are considered to be based on the power ow at the PCC

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
Table 2
Comparison of passive islanding techniques.
Method

Implementation speed

Weakness

Improvement

UFP/OFP
UVP/OVP

Easy but reaction time

unpredictable and variable

Large NDZ

Compared with P-V and P-Q, for

constant current controlled
inverters

PJD

Difcult in implementation and

hard to choose threshold

failed to detect islanding when DG

power generation matches the
power demand of local load

Controlled by using a PLL

THD

Easy but hard to choose threshold

failed to detect islanded in case of

low distortion of voltage and
current output of inverter or high
quality load

Voltage unbalance

between the utility and PV inverter, which refers to the active power
(P) and reactive power (Q).
However, the primary weakness of the UVP/OVP and UFP/OFP
is the large NDZ. Therefore, some improvement was made to
overcome the large NDZ components. A method is proposed to
reduce the NDZ of UVP/OVP and UFP/OFP by comparing the PV
and PQ characteristics of the controlled constant current inverters [23]. The islanding detection based on the performance of the
interface control, which is an additional parameter, was implemented in parallel to the UVP/OVP and UFP/OFP to reduce
the NDZ.

2.2.1.2. Voltage phase jump detection. Phase jump detection (PJD)

requires monitoring the phase difference between the terminal
voltage and output current of the inverter for a sudden jump
[24]. If islanding occurs, the inverter and local load depart from the
main system. The PJD technique searches for a rapid change in
phase angle to detect islanding. This method has easy
implementation because only modifying the phase locked loop
(PLL) required by the inverters for utility synchronization is
needed. The capability to deactivate the inverter is only required
when the phase errors exceed some threshold. The method also
does not affect the power quality of the inverter and can be used in
multiple inverter systems [12]. However, PJD is unsuitable in
detecting islanding for all the operating conditions of the load
[1]. The drawback of the PJD leads to a failure in detecting the
islanding condition when the DG power generation matches the
power demand of the local loads. Nevertheless, the NDZ of the PJD
method is smaller than standard over/under protection devices
because of the dependence of the PJD method on the only the
power factor. This method can be implemented in the inverter
using a PLL either in analog or digital forms [25].

2.2.1.3. Harmonics measurement. The measurement of the harmonics

(line voltages and currents) is known as the detection of harmonics
[26,27]. This method monitors the change of the total harmonic
distortion (THD) at the PCC [25]. If the THD exceeds a dened
threshold, the inverter should disconnect the DGs. However,
selecting a trip threshold, which does not lead to nuisance tripping
of the inverter, is easy because the distortion level rapidly changes as
the nonlinear load switches on and off [28]. Two monitoring
parameters, namely, voltage unbalanced and the THD of current, are
also proposed to detect the islanding operation of DG [29]. However,
this method is difcult because of the high Q factor detection and the
threshold selection problem [5].

Not applicable to signal phase

system

Combine two or more methods,

e.g. Unbalanced voltage and THD
of current and voltage
magnitude variation

2.2.1.4. Voltage unbalance. The voltage unbalance (VU) generally

varies because of the topology changes of the networks and the
load despite the small changes in the DG loading [3,4,29].
Therefore, effectively detecting the islanding operation is
possible if the unbalanced of the three-phase output voltage of
the DG is continuously monitored.
Special identication was conducted to monitor the VU deviation from the steady state and normal loading conditions. Therefore, each step is compared with the VU identication values and
any abrupt VU is identify as islanding.
Jang et al., dened the one-cycle average of the voltage
unbalanced and voltage unbalanced variation, which was checked
at every 1/4 cycle (4.17 ms) [29].
The survey shows that most of the preferences in choosing
passive islanding detection are the cost and simplicity of implementation. Passive techniques are effective for majority of the
situation disturbances that occurr in the grid [10].
However, the major drawback of passive techniques is the large
NDZ, which fails to detect the islanding condition. The local load
affects the determination of the islanding detection. Therefore,
those limitations can be answered by active techniques, which are
categorized in the next section. Table 2 summarizes the existing
major issues of the passive technique method with their advantages and disadvantages.
2.2.2. Active techniques
Active techniques have recently been applied by introducing a
small disturbance to grids, which is the response of the intern with
the grid and deciding if the grid is in the islanding condition
[3,4,10,28,30]. Various active islanding detection with nonarticial intelligent techniques exist, some of which are described
and discussed in detail below.
2.2.2.1. Impedance measurement. The impedance measurement
method is the same as the passive technique, which measures
the system impedance changes caused by islanding. In an active
direct method however, a shut inductor is momentarily connected
across the supply voltage from time to time, and the short circuit
current and supply voltage reduction are used to calculate the
power system source impedance [3,4]. A large number of
impedance detection methods have recently been proposed
because of the belief that this method has no NDZ in the singleinverter case.
Therefore, the experimental study veries the impedance
detection test based on the islanding detection in a singleinverter case [31].

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
The experiment reveals that the impedance detection method
based on islanding detection does have a NDZ in the singleinverter systems. Given the parallel RLC load however, high-Q
loads cause the most trouble for impedance detection such as with
frequency-shifting islanding detection methods. Ensuring that the
inverter maximum power point is enabled with impedance
detection is important because impedance detection can be a
reliable anti-islanding method even for a small duty ratio value.
The effectiveness of impedance detection in the single-inverter
case is greatly enhanced by the addition of a time-varying phase
shift. This improvement comes with an additional cost from the
small amounts of sub-harmonics in the PV inverter output. Therefore, other active techniques can be used to surpass the limitation
of impedance measurement.
2.2.2.2. Slip-mode frequency shift (SMS). SMS is one of the three
methods described in this study that uses positive feedback for
islanding detection, namely, amplitude, frequency, and phase.
The method applies positive feedback to the phase of the
voltage as a method in shifting the phase and, subsequently, the
short-term frequency. The SMS is used to detect the islanding
condition because of the easy implementation of the method
caused by the involvement of only a slight modication of a
required component. SMS is also highly effective in islanding
prevention (small NDZ) compared with other active techniques
[23].
However, SMS causes system-level power quality and transient
response problems at very high penetration levels and feedback
loop gains. This problem is common to all three methods that
utilize positive feedback.
SMS also introduces a phase shift perturbation, which can lead
to noise, measurement inaccuracy, and quantization error in
practice. This limitation can be answered by introducing an
additional phase shift called the improved-SMS (IM-SMS) [32].
The IM-SMS was veried through digital simulation and experimentation, which result simplicity, easy implementation, and high
reliability.
2.2.2.2.1. Active frequency drift (AFD). The principle of the active
frequency drift (AFD) or frequency bias method is forcing variations in the inverter output using positive feedback to accelerate
the frequency of the inverter current[12,33].
The AFD uses the waveform of the inverter current, which is
injected into the PCC. The waveform drifts of the grid is not
present as a stabilizing inuence [5,12,33,34]. The advantage of
AFD is the ease of implementation in microcontroller-based
inverters. However, all inverters must have identical AFD or else
would fail to detect islanding conditions in the multiple inverter
case [5].
AFD is also only effective for purely resistive loads [35]. Therefore, the automatic phase-shift method (APS) is proposed to
mitigate the non-detection problem with SMS and AFD. Only the
starting angle of the inverter output current is changed using APS
based on the frequency of the inverter terminal voltage [36].
However, problems occur in islanding detection such as the
assumption of the unity-power factor operation of the PV

inverters, which fails in the frequency-shift method or APS, and

the unwarranted effectiveness [36].
The rms value and Fourier series coefcients of the current
waveform are used to improved the conventional AFD. The
operational characteristics of the method are also analyzed, which
gave less total harmonic than conventional AFD. This method can
reduce about 30% of the THD in the current waveform. Therefore,
the method can more rapidly detect the islanding with improved
NDZ. However, the method can be further improved using the
active frequency drift with positive feedback (AFDPF) [37].
2.2.2.2.2. Sandia frequency shift (SFS). The SFS method, normally known as AFDPF, is an extension of the AFD and is another
method that uses positive feedback for islanding detection [25].
SFS is not difcult to implement and has one of the smallest
NDZs of all the active islanding detection techniques. SFS has been
implemented in combination with Sandia Voltage Shift (SVS)based islanding detection. This combination is extremely effective
[5]. However, SFS has issues in power quality and system stability,
which can cause an undesirable behavior in the system response.
The drawback of AFD and AFDPF, which is the degree of power
quality caused by the discontinuous waveform, can be overcome
using the proposed new phase drift anti-islanding algorithm [19].
In the proposed method, the phase difference of the terminal
voltage and current is applied to detect the islanding condition.
The method use a current command by increasing the phase
difference either k2 or  k2 when the islanding phenomena occurs.
Therefore, this method can reduce the NDZ and minimize the
deterioration of power quality.
SFS is also used to detect the deterioration of the islanding
condition using multiple PV inverters. The performance of the SFS
method is notably dependent on the optimal design of its parameter. Therefore, a mathematical formula is derived to determine
the optimal setting of the SFS islanding detection parameters with
a multi inverter-based DG, such as PV systems, to eliminate the
NDZ condition.
2.2.2.2.3. Sandia voltage shift (SVS). The third positive feedbackbased method of islanding detection is the SVS, which applies
positive feedback to the amplitude of the voltage. SVS is the most
effective in islanding prevention among the three positive
feedback-based methods. However, SVS may produce minimal
effects on the utility system transient response and power quality.
An active technique was used by adding the disturbance
current through the voltage-source control (VSC) to effectively
detect the islanding condition. The variation of the time-varying
amplitude of the PCC voltage is also determined when the
disturbance remains constantly unchanged [20]. This method has
a faster detection speed and a smaller NDZ than the normal
technique if the suitable accelerating factor is selected.
Karimi et al. suggested a new islanding detection method based
on a negative-sequence current injected using VSC [38]. The
method detects and measures the negative-sequence voltage at
the PCC of the VSC using a unied three-phase signal processor,
which is an improved PLL system that gives a high degree of
immunity to noise. Therefore, the method allows for islanding
detection by injecting a small negative-sequence current (2% to

Table 3
Comparison of active islanding detection technique.
Method

Implementation and speed

Impedance measurement
SMS
AFD
SFS
SVS

Easy and fast

Medium and slow
Easy and medium
Difcult and relatively fast
Medium and fast

Weakness

Ineffective under certain load, e.g. RLC resonant load

Problem in power quality, system stability
Increased harmonic distortion

NDZ
Large for high Q load
Large for high Q load
Large for high Q load
Exist for high Q load but less compared to others
Very less

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
3%). The simulation reveals that the islanding event can be
detected within 60 ms under the UL1741 anti-islanding test condition [39].
Bahrani et al. also presented a solution in obviating the NDZ of
an existing islanding detection method to prove the method had
NDZ in the load parameter space. The method was tested using a
multi-DG system and was veried for islanding detection. The
experiment revealed that the islanding events can be detected
within 120 ms in the worst case scenario because the load
parameter is in the NDZ space [39].
The survey shows that an active technique based on islanding
detection can defeat the passive technique limitation by giving a
small NDZ. However, the active method has better reliability than
the passive technique. The main limitation of the active technique
is the perturbation in the system. The detection time for the
islanding detection is slow because extra time is needed for the
system to response to the perturbation. Table 3 summarizes the
existing major issues when using the active technique method.
2.2.3. Hybrid technique
The hybrid technique is a combination of the active and passive
techniques. The active technique is applied only if islanding is
detected based on the passive technique. Some of the hybrid
techniques that do not utilize articial intelligence are briey
discussed.
2.2.3.1. Voltage unbalance and frequency set point. Menon and
Nehrir presented a hybrid method based on the positive
feedback (PF) and the voltage unbalance and total harmonic
distortion (VU/THD) techniques, where in the drawback of both
techniques are canceled out when simultaneously applied [40].
The calculation of the VU for each of the DG is used instead of the
THD because the VU is more sensitive to disturbance than THD.
Therefore, any disturbance applied to DGs could produce a spike in
the VU. The technique also efciently discriminates between the
load switching and islanding condition.
2.2.3.2. Technique based on voltage and real power shift. This
technique uses an average rate of the voltage change (passive
technique) and a real power shift (active technique) to answer the
limitations of the active and passive techniques in islanding
detection [35]. The technique can detect the islanding condition
with multiple DG units operating at a unity power factor.
However, the real power shift (RPS) is applied to the system if
the passive technique cannot perfectly detect the islanding condition. RPS can eliminate the injecting disturbance from time to
time to detect islanding similar to other active techniques [35].
However, RPS only can change the real power of the DG at the
unity power factor. In the proposed method, only one DG changes
the real power compared with the positive feedback technique,
where all the DGs work together to inject perturbation in the
system [35].
2.2.3.3. Voltage uctuation injection. This technique is based on the
voltage uctuation injection, which can be obtained using a high
impedance load [41]. The islanding detection correlation factor
(CF) is proposed for small-scale DGs, typically less than 1 kW. The
two-stage method, which is the passive technique (rate of change
of frequency (ROCOF)/rate of change of voltage (ROCOV)) for the
protection scheme, and the active technique (CF) as a backup is
applied to attain higher effectiveness. This technique uses digital
signal processing to calculate the ROCOF, ROCOV, and CF of the
distribution synchronous generator and to accurately decide
between islanding and non-islanding disturbances [41].

However, ROCOF is used as a protection scheme, whereas the

detection of the output active power variation and reserve VAR
detection is used as a backup protection scheme in islanding
detection [8]. The NDZ is reduced when the ROCOF is applied with
a change in the active technique for islanding detection. The
technique can also reduce the insensitivity to the system parameter variation.
2.2.3.4. Hybrid SFS and Qf islanding technique. The SFS and Qf
curve are proposed based on islanding detection to improve the
SFS and reduce the NDZ [42]. The optimum SFS is calculated using
the analytic formula, where the bacterial foraging algorithm is
being used to search the optimal gain of the SFS to eliminate NDZ.
The Qf droop curve technique is then added to increase the
effectiveness of the SFS-based islanding detection.
2.3. Signal processing methods used for identication
of island mode operation
The islanding condition must be detected as quickly and
accurately as possible. Therefore, researchers are currently looking
into intelligent methods in detecting and classifying the islanding
condition. Signal processing techniques have aided researchers in
understanding the existence of an island mode operation regardless if the location of control is local or central. The implementation of the signal processing technique allows for the extraction of
the hidden features of the measured signals to detect the islanding
condition. These extracted features can then serve as inputs to the
articial intelligent (AI) classier to perform the classication of
the islanding and non-islanding detection. Common AI classiers
used in islanding detection include the decision tree (DT), rulebased techniques, articial neural network (ANN), probabilistic
neural network (PNN), fuzzy logic (FL), and support vector
machines (SVM).
The signal processing methods present various features, such as
time-frequency distribution (TFD) of a time series, which eases the
analysis and quantication of the signals regardless of the succeeding classication technique shown in Fig. 4. The linear TFD
techniques are popularly implemented in determining the islanding condition because such implementation is faster than those of
non-linear methods. Fig. 4 shows the steps involved in performing
the islanding condition classication. The following sections
describe some of the SP tools used in islanding detection. The
islanding detection techniques that use signal processing and the
succeeding AI classiers are discussed further in the following
subsections.
2.3.1. Wavelet-transform (WT)
The wavelet theory is the mathematical model for nonstationary signals with a set of components in the form of small
waves called wavelets [43]. The original wavelet is generated from
one original wavelet called the mother wavelet, which is then
further extended to allow the wavelet to analyze the nonstationary signals in the frequency band. WT can be either
continuous (CWT) or discrete (DWT). Using a wavelet is advantageous because the wavelet does not need to assume the stationary
or periodicity of signal. The WT is capable simultaneously comprehending time and frequency information because wavelet has
long windows at low frequencies and short windows in high
frequencies. Therefore, WT can supervise the discontinuities and
transients in time-varying signals to enhance islanding detection
studies.
CWT is applied using an online measuring technique for the
voltage in DG units [44]. The Mallat decomposition algorithm was
used in extracting the noise in the signals to eliminate the noise.

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.

PARAMETERS
SIGNALS
(VOLTAGE,
CURRENT,
FREQUENCY)

DATA
ACQUISITION
SYSTEMS

DISTURBANCE
DETECTION
USINGSIGNAL
PROCESSING

FEATURES
EXTRACTION

TYPES OF
DISTURBANCES

CLASSIFICATION
TECHNIQUE

DECISION
MAKING
TECHNIQUE

Fig. 4. Basic blocks of intelligent islanding detection technique and classication.

The technique creates numerous coefcients to reduce the computational efciency of the algorithms, which allows researchers
to focus more on using the DWT technique instead of the CWT
technique.
Pigazo et al. proposed a WT method to obtain the time
localization of signals from a single-phase PV system [45,46].
The method used ve decomposition levels for the islanding
operation mode, and Biorthogonal 1.5 was implemented for the
anti-islanding detection algorithm. The advantage of the technique
is the reduction of the number of sensors needed and the
minimized computational burden of the anti-islanding algorithm.
Hsieh et al. proposed a WT-based approach to monitor voltage
and frequency variations, where the Daubechies wavelet served as
the basis [47]. Some of the useful features in this method are the
improvement in the islanding detection capability of protection
relay, the simultaneous observation of information, and the
simplicity to programming. The technique can give feasibility,
exibility, and robustness to the proposed method as tested in
several scenarios.
In [48], WT was applied to a negative sequence voltage and
current signals. The change in energy and standard deviations
coefcients were then used to distinguish between islanding and
non-islanding events. Daubechies db4 was used for islanding
detection in this technique. The islanding can clearly be detected
for one cycle signal data using rst level wavelet coefcients (d1)
of the energy and standard deviations. DWT with Daubechies db4
was similarly used to reduce the NDZ to zero. Db4 is used because
of its compacted and localization properties [49]. Db4 also utilizes
the spectral changes in higher frequency components by applying
second level wavelet coefcients (d2), which are more robust and
less affected by noise.
One of the limitations of using DWT is the merging of the
frequencies, especially at high frequencies. Therefore, the wavelet
packet transform (WPT) based on voltage and current signal at PCC
is introduced in [50]. The method is based on the rate of change of
power (ROCOP) from the DG power output. The method introduced a new index called the node rate of change of the power
(NROCOPI), which is used to quantify the change of the power at
each WPT sub-band. Daubechies of order 10 served as a basis of
WPT, which had a smaller number of wavelet coefcienta compared with others without affecting the accuracy of the results.
In [51], the wavelet was applied to detect the islanding
condition of wind turbines. The method used DWT with Daubechies db5 to determine the best wavelet basis function with
accurate results. However, all these methods necessitated the
measurement of several electrical quantities, which requires much
time in detecting islanding. Therefore, the wavelet was used to

detect islanding using only the terminal current of DG as a

parameter in [52]. The Haar mother wavelet was used to detect
the islanding event because this wavelet requires the least
decomposition levels and, therefore, the least detection time for
islanding detection. In [53], WT based on detecting the standalone operation of DG systems was introduced.

2.3.2. S-transform
The major drawback of WT is the inability to detect islanding
conditions under a noisy condition. Therefore, a modication of
WT called S-Transform was introduced [5456]. S-transform
produces a time-frequency representation of a time series and
has a frequency dependent resolution that simultaneously localizes the real and imaginary spectra. S-transform also provides
multi-resolution while retaining the absolute phase of each
frequency component, which is useful in detecting disturbances
in the presence of noise.
In [55,56], the S-transform was used to extract the negative
sequence voltage during an islanding event. The energy content
and standard deviation of the S-transform contour was clearly
shown in detecting islanding events and disturbance because of
load rejection. In [57], the negative sequence voltage and current
processed through the S-transform and spectral energy content
was calculated. Therefore, the previous techniques of the Stransform clearly required the energy content to detect the
islanding condition. The drawback of S-transform is the need for
more computation time and memory to process the signal than
other techniques. Table 4 summarizes the signal processing
techniques in islanding detection.
2.4. Intelligent classiers
The information generated by any type of signal processing
method can be used to identify the existence and characteristics of
the island mode operation. In this section, the use of intelligent
classiers are investigated.

2.4.1. Articial neural network (ANN)

The articial neural network (ANN) is a computational structure model of a biological process that attempts to implement the
mathematical model instead of using a biological brain neural
network in which the brain contains all the useful information and
the data memory [58]. This model has numerous interesting and
attractive features that can be used to identify any changes in the
data. Therefore, the model is widely used in numerous areas,

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
Table 4
Comparison of SP in islanding detection at distributed generation.
References

SP method

Parameter

Key indicator

Islanding detection (time

detection)

Zhu[44]

CWT (Wavelet)

Voltage, Current, Power

Modulus local maxima

Pigazo [45,46]
Samantaray [48]

DWT (Wavelet)
DWT (Wavelet)

Hanif [49]

DWT (Wavelet)

Karegar [51]
Shariatinasab [52]

Wavelet Packet Transform

(WPT)
DWT (Wavelet)
DWT (Wavelet)

Energy value of second level

wavelet coefcient
NROCOPI

2.5 power frequency

cycles

Morsi [50]

Voltage, Current, Power

Voltage and current signals.
Change in energy and
standard deviation
Spectral change of PCC
voltage
Voltage and current at PCC

0.6 s (rate of change of

frequency)
Less than 2 s (45 ms)

Voltage
Terminal current of DGs

Fifth decomposition level.

third decomposition level

Less than 0.2 s

One third of cycles (less
than 5.5 ms for 60 Hz)

Ray [54]
Ray [55], [56]

S-transform
S-transform

Voltage
Voltage

Coherency
Energy and standard
deviation indices

including in islanding detection [5966]. The ANN is usually used

with a signal processing technique such as WT.
Yin presented the combination of the signal processing and the
ANN classier for islanding detection [28]. The output voltages of
the inverter were sampled, and the signal frequency domain was
obtained using FFT. Immunological principles were then used,
where two modules (T-module and B-module) were developed to
respond to inverter islanding. The T-module was used to detect the
islanding conditions, and the B-module played role in improving
the detection coverage space. The experiment and data analysis
revealed that the algorithm can effectively detect the islanding
condition of the inverter.
In a related work, a hybrid technique was approached using
ANNs in detecting the islanding condition [67]. This ANN acted like
machine learning technology in processing and analyzing the large
set of data from the simulation. This AI compensator was
embedded in an automatic voltage regulator (AVR) control loop.
In [8], the ANNs were combined with WT, which is capable of
decomposing the signals into different frequency bands. The
features are then trained using the ANN model to identify the
islanding condition. The approach can detect islanding conditions
with a high degree of accuracy and high-quality factor of load
performance. This technique is also suitable for multiple DG
applications.
ANNs were also combined with the wavelet to detect islanding
[68]. The DWT was used to extract the feature from current
signals. The signal was then extracted using a correlation coefcient and was validated using ANN. Only one signal was analyzed,
and only one of the ANN input is used to detect the islanding
condition. This algorithm showed high classication efciency,
which makes the algorithm suitable for a practical system.

2.4.2. Probabilistic neural network (PNN)

Another classication technique that can be used is the
probabilistic neural network (PNN), which is based on a Bayesian
classier technique commonly used in classical patternrecognition applications [69]. PNNs contain four layers, which
are the input layer, pattern layer, summation layer and output
layer, where each layer has its own function in classifying the
features [70]. A PNN does not require a learning process.
The DWT integrated with PNN was proposed for the classication technique in [71]. The combination used the multi-resolution
analysis (MRA) of the DWT and the Parsevals theorem to extract
energy distribution features at different resolution levels. The
features were then classied using the PNN. In [72], PNN were

d1 Coefcients

also combined with WT to detect islanding. The features from the

wavelet analysis were used as an input of the PNN, and the output
was the class of the fault location. The analysis found that PNN is a
very precise way of classifying the location of the fault.
PNN with various parameters were derived at a target DG
location to detect the islanding condition [73]. The author also
compared other classication techniques such as the DT and radial
basis function techniques.
However, PNN is proven to be more effective in islanding
detection than other techniques. PNN is also proven to reliable
in detecting islanding condition because of the simulation model
and real-time digital simulator test.
In [74], a Bayesian passive technique using ESPRIT was proposed, which is a parametric estimation technique that utilizes
special quantities to estimate a certain parameter of a signal. A
statistical signal-processing algorithm was used to extract the
features from the voltage and frequency at the PCC. A crossvalidation and a test of accuracy of the method was conducted in
several cases to estimate the performance of the proposed
technique in islanding detection.

2.4.3. Decision trees (DT)

Another classication technique is the decision tree (DT). In
[75,76], DT was trained, and the online performance of DT was
evaluated in a controlled islanding strategy. DT was trained in
simulation on a particular pre-fault operating point. The analysis
revealed that the proposed method achieved three main objectives that can be correctly classied for all training cases. However,
limitations, such as the dependence of the threshold on the
splitting criteria for the corresponding DT, were found. This
splitting criterion is also a complex task and can affect the DT
technique for classication problems. The criterion has a misclassication rate of 16.6% for islanding detection, which leads to an
83.3% islanding detection accuracy. In [77], DT was used with a
combination of DWT. The classications extracted the features
from the transient voltage and current signals using the DWT,
where the DT classier was used to identify the islanding condition. The authentication test performed had more than 98%
classication accuracy with a 95% condence level and a response
time of less than two cycles. However, this technique was modied
and applied in hardware as presented in [78]. The band pass lter
was used to replace the function of DWT. The experiment had 94%
accuracy in classifying the islanding and non-islanding and took
14 ms to detect the power island.

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
Table 5
Comprehensive analysis.
References
Yin [28]
Ghazi [59]
Fayyad [9]
Nozahy [60]
Lee [63]
Najy [66]
Arroudi [68]
Lidula [69]
Paul Pham [70]
Lin [71]
Sun [73]
Samantaray [77]

SP method

Classication

Recognition rate (RR)

FFT

ANN
ANN
ANN
ANN
ANN
Nave-bayers
DT
DT, CART
DT
DT
DT
DT, MFs

88.9
95.0
99.0
90.0
100.0
83.33
98.00
94.0
100
98
100

WT
DWT
DWT
ESPIRIT
DWT
DWT

The C4.5 DT was introduced for islanding detection in a DG

system by Lin and Dong [79]. The main advantage of C4.5 DT is the
minimization of the NDZ, the ability to operate under various
operations, and the different network topology. However, the
result from the C4.5 DT was not always the best result because
the technique can be improved with other optimization methods
such as the pruning method of DT. Thomas and Terang similarly
proposed an islanding condition using the DT approach [80]. The
technique obtained the threshold setting by selecting a new set of
parameters, such as current, voltage, active and reactive power,
power factor, and frequency. The parameters were easily measured, and the pattern classication was obtained using DT from
MATLABs. However, the system information of PMU measurements was used with DT classier in [81]. DT was used to detect
abnormal system operation behaviors related to islanding
possibility.
Lidula et al. compared several classication techniques such as
DT, neural network and SVM [82,83] to identify the best classication accuracy in detecting the islanding condition, where the DT is
said to be the best classication technique, followed by the neural
network and SVM.
2.4.4. Fuzzy logic (FL)
The fuzzy logic (FL) technique can also be applied as a
classication technique in islanding detection.
In [84], FL based on three measurement parameters, which are
voltage, ROCOF, and active power derivative (ROCOP), was proposed. The FL was applied only when the situation was uncertain
or unclear in detecting the islanding.
In [85], FL was introduced from the transformation of DT,
where the combination of fuzzy membership functions (MFs)
and the rule base were used to develop the fuzzy rule based. This
proposed technique was easy to implement for online islanding
detection and can handle uncertainties such as noise. The technique proposed was tested with and without the noise and
performed perfectly with 100% detection rate cases in islanding.
However, fuzzy classiers have limitations such as being highly
abstract, the heuristic need for experts for rule discovery, and a
lack of self-organizing and self-tuning mechanisms of NN, which
necessitates exploration of other intelligent techniques. Table 5
summarizes some of the benchmark works in islanding detection.

3. Conclusion
In this paper, an extensive analysis of the various systems of
islanding detection techniques in a power distribution system has
been presented. The necessary features of islanding detection techniques and the potential of local and remote techniques used in the
system have been investigated. Some of the utilities and customer

experiences have been outlined. Various intelligent system applications in islanding detection have also been discussed.
About 85 research publications were classied, discussed, and
appended for quick reference. An overview of the possible techniques used to determine the islanding condition is given and the
improvement made on these techniques are highlighted for the
convenience of readers and to create a broad spectrum. The
overview can be reviewed through annotated bibliographies.

Acknowledgement
This work was supported by Universiti Kebangsaan Malaysia
and the Ministry of Higher Education of Malaysia.
References
[1] Aljankawey, A, Morsi, W, Chang, L, Diduch, C. Passive method-based islanding
detection of renewable-based distributed generation: the issues. In: IEEE
electrical power & energy conference; 2010. p. 18.
[2] Lidula N, Rajapakse A. Microgrids research: a review of experimental microgrids and test systems. Renewable and Sustainable Energy Reviews
2011;15:186202.
[3] Mahat, P, Bak-Jensen, B. Review of islanding detection methods for distributed
generation. In: 2008 third international conference on electric utility deregulation and restructuring and power technologies; 2008. p. 27432748.
[4] Mahat, P, Chen, Z, Bak-jensen, B., Review on islanding operation of distribution
system with distributed generation. In: Power and energy society general
meeting; 2011. p. 18.
[5] Kunte, R, Gao, W. Comparison and review of islanding detection techniques for
distributed energy resources. In: 2008 40th North American power symposium; 2008. p. 18.
[6] Llaria A, Curea O, Jimnez J, Camblong H, Recherche E, Suprieure E, et al.
Survey on microgrids: analysis of technical limitations to carry out new
solutions. Power Electronics and Application (EPE09); 18.
[7] Moradzadeh, M, Rajabzadeh, M, Bathaee, M, Novel, A. Hybrid islanding
detection method for distributed generations. In: Third international conference on electric utility deregulation and restructuring and power technologies; 2008. p. 22902295.
[8] Ezzt, M, Marei, M, Abdel-Rahman, M, Mansour, M., A hybrid strategy for
distributed generators islanding detection. In: IEEE power engineering society
conference and exposition in AfricaPowerAfrica; 2007. p. 17.
[9] Fayyad, Y., Neuro-wavelet based islanding detection technique. In: IEEE
electrical power & energy conference; 2010. p. 16.
[10] Timbus, A, Oudalov, A, Ho, C. Islanding detection in smart grids. In: Energy
conversion congress and exposition (ECCE); 2010. p. 36313637.
[11] Abarrategui, D, Zamora, I. Comparative analysis of islanding detection methods in networks with DG. In: 19th international conference on electricity
distribution Vienna; 2007. p. 2124.
[12] Balaguer-lvarez I, Ortiz-rivera E. Survey of distributed generation islanding
detection methods. IEEE Latin America Transactions 2010;8:56570.
[13] Yin, J, Chang, L, Diduch, C., Recent developments in islanding detection for
distributed power generation. In: Power engineering; 2004. p. 124128.
[14] Chowdhury S, Chowdhury S, Crossley P. Islanding protection of active
distribution networks with renewable distributed generators: a comprehensive survey. Electric Power Systems Research 2009;79:98492.
[15] Ecconnect, assessment of islanded operation of distribution networks and
measures for protection. In: Polymer contents, 18; 2001. p. 830877.
[16] Zeineldin H, Kirtley J. A simple technique for islanding detection with
negligible nondetection zone. IEEE Transactions on Power Delivery
2009;24:77986.
[17] Ropp, M, Larson, D, Meendering, R, Mcmahon, D, Ginn, J, Stevens, J, et al.
Islanding scheme using a commercial automatic meter reading system. In:
IEEE fourth world conference on photovoltaic energy conversion; 2006. p.
23512354.
[18] Zeineldin H, Salama M. Impact of load frequency dependence on the NDZ and
performance of the SFS islanding detection method. IEEE Transaction on
Industrial Electronics 2011;58:13946.
[19] Yoo, C, Jang, D, Han, S, Oh, D, Hong, S. A new phase drift anti-islanding method
for grid-connected inverter system. In: Eighth international conference on
power electronicsECCE Asia; 2011. p. 902906.
[20] Li P, Sheng Y, Zhang L, Yang X, Zhao Y. A novel active islanding detection
method based on current-disturbing. Electrical Machines and Systems
2009;1:15.
[21] Lee S, Park J. New islanding detection method for inverter-based distributed
generation considering its switching frequency. IEEE Transaction on Industry
Applications 2010;46:208998.
[22] Yu B, Matsui M, Yu G. A correlation based islanding detection method using
current magnitude disturbance for PV system. IEEE Transaction on Industrial
Electronics 2010;58:293543.

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
[23] Zeineldin H, El-Saadany E, Salama M. Impact of DG interface control on
islanding detection and nondetection zones. IEEE Transactions on Power
Delivery 2006;21:151523.
[24] Hu, W, Sun, Y. A compound scheme of islanding detection according to
inverter. In: Asia-Pacic power and energy engineering conference; 2009. p.
14.
[25] Teoh W, Tan C. An overview of islanding detection methods in photovoltaic
systems. World Academy of Science Engineering and Technology
2011:67482.
[26] Khamis A, Armstrong M, Sulaiman M, Rahman A. Location of embedded
generation installation at long distribution line feeder due to harmonic
performances. Distribution 2009;10:2633.
[27] Khamis, A, Armstrong, M, Sulaiman, M. The impact of embedded generation
due to harmonic performance. In: 2009 Third Asia international conference on
modelling & simulation; 2009. p. 514519.
[28] Yin, GA. Distributed generation islanding detection method based on articial
immune system. In: IEEE/PES transmission & distribution conference &
exposition: Asia and Pacic; 2005. p. 14.
[29] Jang S, Kim K. An islanding detection method for distributed generations using
voltage unbalance and total harmonic distortion of current. IEEE Transaction
on Power Delivery 2004;19:74552.
[30] Bae B, Jeong J, Lee J, Han B. Islanding detection method for inverter-based
distributed generation systems using a signal cross-correlation scheme.
Journal of Power Electronics 2010;10:7628.
[31] Roop, M, Ginn, J, Stevens, J, Bower, W, Gonzales, S. Simulation and experimental study of the impedance detection anti-islanding method in the singleinverter case. In: Proceedings of fourth world conference. IEEE photovoltaic
energy convers, 2; 2006. p. 23792382.
[32] Liu F, Kang Y, Zhang Y, Duan S, Lin X. Improved SMS islanding detection
method for grid-connected converters. IET Renewable Power Generation
2010;4.
[33] Hanif M, Street K, Gaughan K. A discussion of anti-islanding protection
schemes incorporated in a inverter based DG. Environment and Electrical
Engineering (EEEIC); 15.
[34] Kotsopoulos, A, Duarte, J, Hendrix, M, Heskes, P., Islanding behaviour of gridconnected PV inverters operating under different control schemes. In: Power
electronics specialist conference (PESC02); 2002. p. 15061511.
[35] Mahat P, Chen Z, Bak-jensen B. A hybrid islanding detection technique using
average rate of voltage change and real power shift. IEEE Transactions on
Power Delivery 2009;24:76471.
[36] Hung G, Chang C, Chen C. Automatic phase-shift method for islanding
detection of grid-connected photovoltaic inverters. IEEE Transaction on
Energy Conversion 2003;18:16973.
[37] Yafaoui A, Wu B, Kouro S. Improved active frequency drift anti-islanding
detection method for grid connected photovoltaic systems. IEEE Transaction
on Power Electronics 2012;27:236775.
[38] Karimi H, Yazdani A, Iravani R. Negative-sequence current injection for fast
islanding detection of a distributed resource unit. IEEE Transactions on Power
Electronics 2008;23:298307.
[39] Bahrani B, Karimi H, Iravani R. Nondetection zone assessment of an active
islanding detection method and its experimental evaluation. IEEE Transactions on Power Delivery 2011;26:51725.
[40] Menon V, Nehrir M. A hybrid islanding detection technique using voltage
unbalance and frequency set point. IEEE Transactions on Power Systems
2007;22:4428.
[41] Chang, WA. Hybrid islanding detection method for distributed synchronous
generators. In: The 2010 international power electronics conference; 2010. p.
13261330.
[42] Vahedi, H, Noroozian, R, Jalilvand, A, Gharehpetian, G., Hybrid SFS and Qf
islanding detection method for inverter-based DG. In: IEEE international
conference on power and energy (PECon2010); 2010. p. 672676.
[43] Polikar, R. The story of wavelets 1. In: IEEE proceeding CSCC99; 1807. p. 16.
[44] Zhu Y, Yang Q, Wu J, Zheng D, Tian Y. A novel islanding detection method of
distributed generator based on wavelet transform. Electrical Machines and
Systems (ICEMS 2008); 26868.
[45] Pigazo A, Moreno V, Liserre M, Aquila A, Bari P. Wavelet-based islanding
detection algorithm for single-phase photovoltaic (PV) distributed generation
systems. Elettrotecnica; 240913.
[46] Pigazo A, Liserre M, Mastromauro R, Moreno V, Aquila A. Wavelet-based
islanding detection in grid-connected PV systems. IEEE Transaction on
Industrial Electronics 2009;56:444555.
[47] Hsieh C, Lin J, Huang S. Enhancement of islanding-detection of distributed
generation systems via wavelet transform-based approaches. Electrical Power
& Energy Systems 2008;30:57580.
[48] Samantaray, S, Pujhari, T, Subudhi, B., A new approach to islanding detection
in distributed generations. In: Third international conference on power
systems; 2009. p. 2530.
[49] Hanif M, Dwivedi U, Basu M, Gaughan K. Wavelet based islanding detection of
DCAC inverter interfaced DG systems. UPEC2010; 37.
[50] Morsi, W, Diduch, C, Chang, L., A new islanding detection approach using
wavelet packet transform for wind-based distributed generation. In: Second
IEEE international symposium on power electronics for distributed generation
systems; 2010. p. 495500.
[51] Karegar H, Sobhani B. Wavelet transform method for islanding detection of
wind turbines. Renewable Energy 2012;38:94106.

[52] Shariatinasab, R. New islanding detection technique for DG using discrete

wavelet transform. In: IEEE international conference on power and energy
(PECon2010); 2010. p. 294299.
[53] L, J, Lui, Y, Luor, T, Huang, S. Method and system for detecting stand-alone
operation of a distributed generating system ; 2008.
[54] Ray, P, Mohanty, S, Kishor, S, Dubey, H., Coherency determination in gridconnected distributed generation based hybrid system under islanding
scenarios. In: IEEE international conference on power and energy
(PECon2010); 2010. p. 8588.
[55] Ray, P, Kishor, N, Mohanty, S. S-Transform based islanding detection in gridconnected distributed generation based power system. In: IEEE international
energy conference; 2010. p. 612617.
[56] Ray P, Mohanty S, Kishor N. Disturbance detection in grid-connected distributed generation system using wavelet and S-Transform. Electric Power
Systems Research 2011;8:80519.
[57] Samantaray S, Samui A, Babu B. S-Transform based cumulative sum detector
(CUSUM) for islanding detection in distributed generations. Power Electronics,
Drives and Energy Systems (PEDES) 2010:16.
[58] Fukuda T, Shibata T. Theory and applications of neural networks for industrial.
IEEE Transaction on Industrial Electronics 1992;39:47288.
[59] Esen H, Ozgen F, Esen M, Sengur A. Articial neural network and wavelet
neural network approaches for modelling of a solar air heater. Expert Systems
with Applications 2009;36:112408.
[60] Esen H, Inalli M, Sengur A, Esen M. Performance prediction of a groundcoupled heat pump system using articial neural networks. Expert Systems
with Applications 2008;35:19408.
[61] Irani R, Nasimi R. Application of articial bee colony-based neural network in
bottom hole pressure prediction in underbalanced drilling. Journal of Petroleum Science and Engineering 2011;78:612.
[62] Choi J, Oh H, Lee H, Lee C, Lee S. Combining landslide susceptibility maps
obtained from frequency ratio, logistic regression, and articial neural network models using ASTER images and GIS. Engineering Geology
2012;124:1223.
[63] Millie D, Weckman G, Young W, Ivey J, Carrick H, Fahnenstiel G. Modeling
microalgal abundance with articial neural networks: demonstration of a
heuristic Grey-Box to deconvolve and quantify environmental inuences.
Environmental Modelling & Software 2012;38:2739.
[64] Pahlavan R, Omid M, Akram A. Energy inputoutput analysis and application
of articial neural networks for predicting greenhouse basil production.
Energy 2012;37:1716.
[65] Isik S, Kalin L, Schoonover J, Srivastava P, Lockaby B. Modeling effects of
changing land use/cover on daily streamow: an articial neural network and
curve number based hybrid approach. Journal of Hydrology 2013;485:10312.
[66] Oliveira M, Almeida J. Application of articial intelligence techniques in
modeling and control of a nuclear power plant pressurizer system. Progress
in Nuclear Energy 2013;63:7185.
[67] Ghazi R, Lot N. A new hybrid intelligent based approach to islanding
detection in distributed generationl, 22. UPEC; 15.
[68] Elnozahy, M, El-saadany, E, Salama, M. A Robust Wavelet-ANN based technique for islanding detection. In: Power and energy society general meeting;
2011. p. 18.
[69] Parzen E. On estimation of a probability density function and mode. in
statistics. Annals of Mathematical 1962;33:106576.
[70] G AT. Probabilistic neural network for evaluating seismic liquefaction potential. Canadian Geotechnical Journal 2002;39:21932.
[71] Gaing Z. Wavelet-based neural network for power disturbance recognition
and classication. IEEE Transactions on Power Delivery 2004;19:15608.
[72] Othman, M, Amari, H., Online fault detection for power system using wavelet
and PNN. In: Second IEEE international conference on power and energy
(PECon 08); 2008. p. 16441648.
[73] Samantaray S, Babu B, Dash P. Probabilistic neural network based islanding
detection in distributed generation. Electric Power Components and Systems
2011;39:191203.
[74] Najy W, Zeineldin H, Alaboudy A, Woon W. A bayesian passive islanding
detection method for inverter-based distributed generation using ESPRIT. IEEE
Transactions on Power Delivery 2011;26:268796.
[75] Senroy N, Heydt G, Vittal V. Decision tree assisted controlled islanding. IEEE
Transactions on Power Systems 2006;21:17907.
[76] El-arroudi K, Jos G, Kamwa I. Intelligent-based approach to islanding
detection in distributed generation. IEEE Transactions on Power Delivery
2007;22:82835.
[77] Lidula N, Rajapakse A. A pattern recognition approach for detecting power
islands using transient signals Part I: Design and implementation. IEEE
Transaction on Power Delivery 2010;25:30707.
[78] Pham, J, Denboer, N, Lidula, N, Member, G, Perera, N, Rajapakse, A, et al.,
Hardware implementation of an islanding detection approach based on
current and voltage transients. In: Electrical power and energy conference
(EPEC); 2011. p. 152157.
[79] Lin, X, Dong, X, Lu, Y. Application of intelligent algorithm in island detection of
distributed generation. In: Transmission and distribution conference and
exposition (IEEE PES); 2010. p. 17.
[80] Thomas M, Terang P. Islanding detection using decision tree approach. Power
Electronics,Drives and Energy SYStems (PEDES) 2010:16.
[81] Sun, R, Wu, Z, Centeno, V. Power system islanding detection & identication
using topology approach and decision tree. In: 2011 IEEE power and energy
society general meeting; 2011. p. 16.

Edited by Foxit PDF Editor

Copyright (c) by Foxit Software Company, 2004 - 2007
For Evaluation Only.
[82] Lidula, N, Perera, N, Rajapakse, A, Investigation of a fast islanding detection
methodology using transient signals. In: Power & energy society general
meeting (PES09); 2009. p. 16.
[83] Lidula, N, Ajapakse, A. Fast and reliable detection of power islands using
transient signals. In: International conference on industrial and information
systems (ICIIS); 2009. p. 493498.

[84] Rosolowski, E, Burek, A, Jedut, L, New, A. Method for islanding detection in

distributed generation. In: Power.
[85] Samantaray S, El-arroudi K, Jos G, Kamwa I. A fuzzy rule-based approach for
islanding detection in distributed generation. IEEE Transaction on Power
Delivery 2010;25:142733.